MatzeB wrote: So not sure if there is an easy fix for the regression.
We should probably have some logic that scales all BFI values when inlining to better accomodate the new range of frequencies so we loose less precision by accident. Though that won't help with the regression at hand as that would need for `isColdCallSite` to accidentally not work as before to get more inlining... https://github.com/llvm/llvm-project/pull/66285 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
